Does Zoom use your data to train AI? What its terms say, and what still leaves the meeting

The Morse teamPublished November 9, 202611 min read

As of November 2026, Zoom says it does not use your meeting data, meaning your audio, video, chat, screen sharing, and attachments, to train its own AI models or anyone else's. Section 10.2 of Zoom's terms of service, in force since August 11, 2023, says Zoom "does not use any of your audio, video, chat, screen sharing, attachments or other communications-like Customer Content (such as poll results, whiteboard and reactions) to train Zoom or third-party artificial intelligence models" (Zoom terms). Training isn't the only way your words travel, though. Zoom's AI features send meeting content to models from OpenAI, Anthropic, and Google to produce summaries and answers (Zoom whitepaper). Feedback you choose to send with a prompt can be used to improve those features (Zoom).

We read Zoom's terms of service, its privacy statement (last updated September 17, 2026), five pages of its AI whitepaper, eight of its support articles, its June 2026 ZoomMate announcement, its ZoomMate pricing page, and archived copies of its 2023 terms and blog post, all on November 9, 2026. Zoom deserves credit in two places. It names the outside AI companies it uses, and it offers an admin setting that produces a meeting summary without keeping the transcript.

Zoom AI privacy settings at a glance

Read November 9, 2026, for Zoom Workplace accounts.

What does Zoom's terms of service say about AI?

It says no to training, and it hasn't changed since 2023. The current terms still carry an effective date of August 11, 2023 (Zoom terms). The no-training sentence sits in bold in section 10.2, under a list of five "Permitted Uses" for your content, such as providing the service and following Zoom's privacy statement. The update note at the bottom says Zoom narrowed "the scope of Zoom's licenses" in that version.

The promise covers AI prompts too. Zoom's page on how its AI features handle data says the commitment "includes the content mentioned in the chart above," and that chart lists "user inputs" among the content ZoomMate and the AI chat panel use (Zoom). Zoom's privacy statement repeats the same promise (Zoom privacy statement).

One kind of data sits outside the promise. Section 10.5 of the terms calls telemetry, product usage, and diagnostic data "Service Generated Data" and says Zoom owns it outright (Zoom terms). The privacy statement lists clicks, keystrokes, mute and unmute, and "edits to transcript text, where authorized by the account owner" as usage information Zoom collects (Zoom privacy statement).

How Zoom's AI training wording changed, 2023 to 2026

Most of the top results for this question are news stories from the week of August 7, 2023, which quote Zoom's promise not to train "without your consent." Zoom dropped that wording four days after adding it. Here is each version, from Zoom's own pages and archived copies of them.

Does Zoom send your meetings to OpenAI or Anthropic?

Yes, when you use its AI features, unless your account uses one of Zoom's hosted-model options. Zoom's AI whitepaper lists Zoom-hosted models, Anthropic models, OpenAI models, and Google Gemini models under its "federated approach" (Zoom whitepaper). It also names Perplexity for some web search results and Azure Speech Translation for live audio translation. Zoom's support page gives an example: when you summarize a meeting, the "speech-to-text data" goes "to the relevant model" (Zoom).

Zoom says those providers keep nothing and train on nothing. The whitepaper says Zoom has zero data retention policies with them, "with limited exceptions for trust and safety purposes, such as to detect and report CSAM." Zoom's support page adds, "We do not allow third-party model providers to use your data to improve or train their models" (Zoom). We explain what that kind of agreement does and doesn't cover in what zero data retention means.

Two other setups keep meetings away from the model companies. With "Zoom-Hosted Models Plus," third-party models run on Zoom-managed instances in Amazon Bedrock, "where information is not sent to third-party AI model providers." With "Zoom-Hosted Models Only," Zoom uses only its own models. Zoom uses the first for accounts hosted in the EU or Australia, and other customers can ask Zoom support to switch (Zoom whitepaper).

Can anything you type in Zoom end up improving its AI?

Yes, through feedback. Zoom's feedback guide says "we improve our AI features by using the feedback that you provide, which may include thumbs up/down, preselected text and free text, and context such as user prompts and responses, screenshots, and diagnostic data if submitted" (Zoom). The same page says "Zoom employees may review this feedback."

The context is opt-in. After a thumbs up or down in ZoomMate or the AI chat panel, a checkbox decides whether your prompt and the reply go with it, and an admin has to allow that sharing first. ZoomMate shows the file it will send, contextdata.txt, before you submit. If you'd rather your prompts stay away from Zoom's staff, leave that box unticked, or don't rate replies at all.

How do you turn off Zoom AI Companion?

In the web portal, under Zoom AI. AI Companion isn't the menu's name anymore. On June 24, 2026, Zoom said AI Companion's features are "now natively built into Zoom Workplace" and are "called by what they do now," such as meeting summary and chat compose (Zoom blog). Guides that still send you to an AI Companion tab under Account Settings describe the old menu.

The current path, from Zoom's getting-started guide (Zoom):

1. For yourself: sign in to the Zoom web portal, click Settings, then Zoom AI, and switch off each feature you don't want.
2. For the whole account: an owner or admin clicks Admin Center, then Settings, then Zoom AI.
3. For a group: Admin Center > Users > Groups, then the group's name.
4. To stop people turning it back on: an admin clicks the lock icon next to the setting.

A grayed-out switch means an admin locked it, so only they can change it. Watch the mobile app too. Zoom says a prompt there offers to turn on summaries, and tapping Enable "activates both the meeting summary and auto-start summary settings across all of the user's meetings" (Zoom).

What happened to Custom AI Companion?

ZoomMate replaced it as Zoom's paid AI add-on. Zoom calls ZoomMate "Zoom's new agentic product that builds on Custom AI Companion," and says "Custom AI Companion customers will continue to be supported" (Zoom blog). As of November 9, 2026, ZoomMate costs $20.00 per user a month billed monthly, or $16.67 billed yearly, with 2,200 AI credits per user each month (Zoom pricing). The built-in features, meeting summary included, still come with eligible paid plans at no extra cost.

ZoomMate reaches further into your work than the built-in features. Zoom's whitepaper says it keeps "user-specific memory," can connect to tools like Google Drive, Jira, and CRMs, and saves screenshots from its browser sessions to your conversation history (Zoom whitepaper). Files you upload stay in Zoom File Storage for up to seven days. We found no fixed retention period for ZoomMate or AI chat panel conversations on Zoom's pages; Zoom's guides tell you how to delete them one by one (Zoom).

How can you tell if Zoom AI is on in a meeting?

Look at the top-right corner of the meeting window. When meeting summary or meeting questions are on, "attendees will see an illuminated AI Sparkle icon," and hovering over it shows which features are running and whether a transcript is being kept (Zoom whitepaper). My Notes, Zoom's personal transcription tool, has its own signals. The whitepaper describes a banner saying the meeting is being transcribed, a pencil icon next to that person in the Participants list, and a transcription icon in the corner.

A participant can't switch the summary off. Zoom says "only the meeting host and co-host can start or stop the Meeting summary" (Zoom). A participant can ask the host to stop, and the host can tick Delete meeting assets to remove the summary and transcript made so far. With auto-start on, Zoom says the summary "will begin as soon as someone starts your meeting," whether or not the host is there.

Who gets the Zoom meeting summary?

Whoever the host's sharing setting names, plus the account owner. Zoom's settings page offers four choices under Automatically share summary with, from "Only meeting host" to "All meeting invitees including those outside of our organization" (Zoom). Only people on the original invite list get it. Anyone who joined from a forwarded link doesn't.

From there it can travel. Zoom says "anyone who receives the meeting summary may save and share it with apps and others," and "the account owner can access the meeting summary" (Zoom). Admins can block sharing outside the organization, send a link instead of the text, and auto-delete summaries after a set number of days (Zoom). They can also turn transcripts off and lock them, which Zoom's whitepaper calls zero data retention for meeting summary: the temporary data is "discarded and cannot later be recovered" once the summary is made (Zoom whitepaper).

Bots that other people bring are a separate question, since each follows its own policy. We cover them in is Otter AI private and is Read AI safe.

How Morse keeps your own questions out of a work account

Zoom's training promise holds up on its own pages. On a work account, though, your meeting is stored in an account your employer owns, where the owner can open any summary, the host decides who else gets one, and a rated reply can carry your prompt to Zoom's staff if the box is ticked. Work material belongs in the tools your employer approves, under its rules. The questions you'd rather keep to yourself, like how to respond to something said in that meeting, belong somewhere you hold the account.

On Morse, your chat history is stored in your browser, encrypted, not on Morse's servers (privacy policy). It locks when you sign out and takes two keys to open, one of which never leaves your device. Morse doesn't train AI on your content, sell your data, or show ads, so there's no training switch to hunt for. Each chat has its own privacy mode:

• Private, for the follow-up you'd rather draft alone. It's the default for a new chat. Morse has zero data retention agreements with its inference, search, and other providers, so they keep nothing, and where a provider confirms zero retention on each reply, Morse checks it and refuses a reply that doesn't.
• E2EE, for anything no one else should read. Your browser checks the secure enclave's hardware attestation before anything leaves your device, then encrypts the message so only that enclave can read it. Morse's servers relay ciphertext they can't read. This rests on the enclave hardware and attestation working as designed. E2EE runs open-weight models, without web search or image generation, and comes with Pro, at $19 a month.
• Anonymous, for GPT, Claude, or Gemini without your name on it. Our proxy removes your account, email, and IP address before the message goes out under Morse's own account. The provider sees the message without knowing it's you, and may keep it for a limited time to check for abuse, under its own policy, so leave out names that point to you or your colleagues.

Clearing your browser or deleting your account removes your chats for good, and Morse can't recover them, so keep a copy of anything you need. Because we don't store chats, we usually have very little to give authorities. For how other companies handle training, see which AI companies train on your data, and for Slack, how to stop Slack AI training.

So what should you do about Zoom AI?

• Training: nothing to switch off. Zoom's own pages rule it out for meeting content and AI prompts. Morse doesn't train AI on your content either.
• Feedback: when you rate a ZoomMate or AI chat panel reply, leave the box that shares your prompt and the reply unticked.
• Hosting: check Settings > Zoom AI, and set Automatically share summary with to Only meeting host until you've read the summary.
• Attending: look for the lit sparkle icon, and ask the host to stop it and delete the assets if you didn't agree to it.
• Your own questions: ask them where you hold the account.

Try a chat from the Morse home page without an account, or start free with no card needed.